Advertisers deeply understand the way that existing and potential customers evaluate a purchase of a product or service. This process is typically described as the "purchase funnel," and is a relatively standard way for advertisers to think about approaching their advertising spending against prospective customers. Typically advertisers try to find ways to reach prospects with a message that caters specifically to where that person sits within the purchase funnel.
There are numerous purchase funnel definitions out there. To keep things simple, let's deal with a very basic version:
Awareness > Consideration > Preference > Purchase
Here are a few basic issues to consider when thinking about the purchase funnel verses the advertising inventory that exists in the world. At the opening of the funnel, starting with awareness, there is a vast amount of advertising inventory available, and at a relatively low cost per thousand (CPM) ad impressions. At the end of the funnel, ending with purchase, there is a small amount of ad inventory, and the cost of that inventory is quite high on a CPM basis.
Advertisers also apply two macro methodologies to their advertising spending habits across the entire purchase funnel. They either use direct response approaches to buying ads, or brand approaches to buying ads. To keep it simple, let’s just say that DR advertisers measure to a cost-per-acquisition (a purchase) during the life of the campaign. And let’s call the brand bucket as measuring key performance indicators (KPIs) that are tied to the location and movement of the potential customer through the purchase funnel. This includes elements like brand awareness, brand consideration, brand preference, and some more direct response-like measures like purchase intent.
Large advertisers that spend hundreds of millions, if not billions of dollars annually are also very good at measuring their ROI. You might imagine that if a small percentage of a large budget is spent on measuring advertising effectiveness, they can get pretty comfortable with their return. I’ve heard some consumer products goods companies boast of being able to predict ROI to within a few decimal points of percentage.
While all advertisers will admit that their advertising spend is somewhat inefficient (meaning they don’t exactly know which spend drove which lift), they don’t often talk publicly about their ability to predict results -- as this is highly valuable competitive data. And yet the myth seems to persist that advertisers simply throw spaghetti against the wall and expect half of it to stick -- and therefore are somehow “wasteful” or “unscientific” in their approach to advertising.
The real discussion to have is around how they allocate that spend. Less than 10 percent of all ad dollars spent are allocated to advertising online, and the split between paid search and display ads is pretty even at the aggregate level. Within the display advertising bucket, only a small amount of the total spend is allocated to direct response buying, most dollars are spent on branding.
The above diagram represents inventory as mapped against the purchase funnel (all impressions that exist across all online media could be fit within this triangle) showing the breakdown of buying methodologies (brand and DR) and the relative price paid for the inventory by the advertiser. Note that in the online space, the pointy-end of the funnel is basically paid search advertising. And everything to the left of the vertical dotted line is display advertising, with half of the inventory going to brand advertisers, and half going to DR advertisers.
But note that a proportionally small amount of revenue actually goes against DR spending on display ads. This pool of inventory -- DR display spend -- is made up of inventory sold through contextual networks, display ad networks, and ad exchanges today. The red ball (below) shows revenue from premium brand advertisers -- primarily hand-sold by a human sales force. Over time I predict that we'll see the amount of inventory allocated against DR buyers change as we remove friction from the purchase process by automating the purchase process.
While as a category, paid search will always be high revenue generation per impression (for the search engines), the overall number of impressions that exist are segmented into very small pools of inventory. Each small pool (denoted by a single keyword) tends to be relevant only to a small number of advertisers -- meaning that per advertiser there just isn’t any way to buy very many impressions (even if you’re paying on a cost-per-click).
Keep in mind that while the paid search space boasts more than a half-million advertisers, the display space (across online display and even offline media such as television and print) really is made up of only about 3,000 to 5,000 brands that perform large-scale buys. There is an important take-away from that point. There is not enough inventory in paid search for it to be a large percentage of any one advertiser's budget overall (spreading the entire 4 percent of all spending -- see below -- across hundreds of thousands of advertisers versus online display spreading spend across 3,000 to 5,000 advertisers).
US Ad Spending across media -- 2008
Source: eMarketer (Barclays Capital March 12, 2009) and Eric Picard analysis
*Excludes the Direct Mail and 'Miscellaneous' categories
Paid search simply doesn't have enough viable inventory in any given category to enable any significant advertiser (large spender) to move large portions of their budget to paid search. Take an advertiser who spends $1 billion in the U.S. each year. One could assume that they follow the breakdown that I'm showing as the average in the chart above. However, it is highly unlikely that they are allocating even the 4 percent listed for paid search on this chart out of their total budget. There is simply not enough inventory for them to spend the money on.
People who do not understand advertising in general but do understand the paid search space frequently talk about how all ad dollars will move to paid search over time, "since it is so effective." When you measure your results against CPA during the life of the campaign, very few people will argue against the value of paid search. But when you look at the entire ad spend of a big-budget advertiser, and you look at the myriad KPIs that they use to measure the success of that advertising spend across brand and direct response, it is highly unlikely that paid search will be incredibly important to any one of those advertisers (beyond some retailers ith massive numbers of products they could buy against in an automated fashion, like eBay and Amazon).
Ultimately, paid search matters far more to the search engines than it does to any one advertiser. After all, before you buy a product you have to be aware that it exists, you have to consider whether you need it, you have to weigh your purchase of that product against all the various manufacturers, and then you have to decide where to go and buy it. Paid search certainly can help sell a product if you catch someone further down the purchase funnel, but it takes other methods to get them to that point.
Eric Picard is the advertising technology advisor to the Advertising Platform Engineering team at Microsoft.